Abstract: The classification and recognition of emergency braking behavior based on electroencephalography (EEG) is a key issue in the development of human-centered intelligent assisted driving systems. In order to realize the classification and recognition of emergency braking and normal driving behaviors during driving, a feature representation method based on Phase Locking Value (PLV) was proposed to construct functional brain networks, the feature parameters of significant differences are determined via statistical analysis of the network feature parameters, and the spatial features of EEG were extracted through Log-Euclidean distance. Combined with machine learning algorithm, emergency braking and normal driving behavior are classified and recognized. The results show that the accuracy of emergency braking and normal driving for 17 participants is higher than 84%, and the highest accuracy rate reaches 95.7%, and the analysis of functional brain network results show that in the process of two driving behaviors, the interaction between brain regions involves the whole brain area, and in the emergency braking process, the interaction between brain regions mainly occurs in the frontal-central-temporal lobe area, which is consistent with the brain focusing more on judgment and decision-making under emergency braking. The results of this paper have certain reference value for understanding the dependence between the driver’s corresponding brain zones during driving, especially during emergency braking, and for developing intelligent assisted driving systems to identify emergency braking intentions in advance during driving.